OpenAI expands Daybreak cyber‑security program, IBM joins as partner

On 22 June, OpenAI announced an expansion of its Daybreak program, a defensive cyber‑security initiative that now includes new tools and a specialized model, GPT‑5.5‑Cyber. The model claims to achieve 85.6 % on the CyberGym benchmark and higher scores on ExploitGym and SEC‑bench Pro, aiming to automate the creation of patches for software vulnerabilities at scale.

IBM has entered the OpenAI Daybreak Cyber Partner Program, launching a service that uses the frontier AI models to scan code repositories in read‑only mode, identify and validate vulnerabilities, and assist with patch generation. In parallel, IBM and Red Hat unveiled Project Lightwell, a $5 billion enterprise security clearinghouse focused on securing open‑source software supply chains. Both moves are presented as a response to intelligence warnings that advanced AI is shortening the window between vulnerability discovery and exploitation.